[PATCH v4 00/12] enable nvmet-fc for blktests

[PATCH v4 00/12] enable nvmet-fc for blktests

[Daniel Wagner]

The first version which allows to run blktests in loop without crashing or
hanging forever somewhere. The missing fix was the argument swap patch
for list_add_tail, which ensured no response gets dropped on tbe floor.

Note, this runs with the udev auto connect rule running in the background which
injects a lot of additional noise to the test. For example blktests will warn
about 'device not removed', which is correct but this device was not created by
the test, it was created by the udev rule. nvme/003 fails because it disconnects
two devices instead the exptected one.

These are things we have to address in blktests though. So hopefully with these
patches we are getting close to run blktests on a regular basis.

changes:
v4:
 - dropped patches which got applied
 - dropped 'nvmet-fc: free hostport after release reference to tgtport'
 - reworked commit message of 'nvmet-fc: untangle cross refcounting objects'
   and renamed it to 'nvmet-fc: defer cleanup using RCU properly'
 - added 'nvmet-fcloop: swap the list_add_tail arguments'
   and 'nvmet-fc: use RCU list iterator for assoc_list'
 - added RBs
 
v3:
 - collected all patches into one series
 - updated ref counting in nvmet-fc

v2:
  - added RBs
  - added new patches
  - https://lore.kernel.org/linux-nvme/20230620133711.22840-1-dwagner@suse.de/
  
v1:
  - https://lore.kernel.org/linux-nvme/20230615094356.14878-1-dwagner@suse.de/ 


*** BLURB HERE ***

Daniel Wagner (12):
  nvme-fc: do not wait in vain when unloading module
  nvmet-fcloop: swap the list_add_tail arguments
  nvmet-fc: release reference on target port
  nvmet-fc: defer cleanup using RCU properly
  nvmet-fc: free queue and assoc directly
  nvmet-fc: hold reference on hostport match
  nvmet-fc: remove null hostport pointer check
  nvmet-fc: do not tack refs on tgtports from assoc
  nvmet-fc: abort command when there is no binding
  nvmet-fc: avoid deadlock on delete association path
  nvmet-fc: take ref count on tgtport before delete assoc
  nvmet-fc: use RCU list iterator for assoc_list

 drivers/nvme/host/fc.c       |  47 ++---------
 drivers/nvme/target/fc.c     | 156 +++++++++++++++++++----------------
 drivers/nvme/target/fcloop.c |   6 +-
 3 files changed, 93 insertions(+), 116 deletions(-)

-- 
2.43.0

[PATCH v4 01/12] nvme-fc: do not wait in vain when unloading module

[Daniel Wagner]

The module exit path has race between deleting all controllers and
freeing 'left over IDs'. To prevent double free a synchronization
between nvme_delete_ctrl and ida_destroy has been added by the initial
commit.

There is some logic around trying to prevent from hanging forever in
wait_for_completion, though it does not handling all cases. E.g.
blktests is able to reproduce the situation where the module unload
hangs forever.

If we completely rely on the cleanup code executed from the
nvme_delete_ctrl path, all IDs will be freed eventually. This makes
calling ida_destroy unnecessary. We only have to ensure that all
nvme_delete_ctrl code has been executed before we leave
nvme_fc_exit_module. This is done by flushing the nvme_delete_wq
workqueue.

While at it, remove the unused nvme_fc_wq workqueue too.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/host/fc.c | 47 ++++++------------------------------------
 1 file changed, 6 insertions(+), 41 deletions(-)

diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c
index e2308119f8f0..7006f4caac2f 100644
--- a/drivers/nvme/host/fc.c
+++ b/drivers/nvme/host/fc.c
@@ -221,11 +221,6 @@ static LIST_HEAD(nvme_fc_lport_list);
 static DEFINE_IDA(nvme_fc_local_port_cnt);
 static DEFINE_IDA(nvme_fc_ctrl_cnt);
 
-static struct workqueue_struct *nvme_fc_wq;
-
-static bool nvme_fc_waiting_to_unload;
-static DECLARE_COMPLETION(nvme_fc_unload_proceed);
-
 /*
  * These items are short-term. They will eventually be moved into
  * a generic FC class. See comments in module init.
@@ -255,8 +250,6 @@ nvme_fc_free_lport(struct kref *ref)
 	/* remove from transport list */
 	spin_lock_irqsave(&nvme_fc_lock, flags);
 	list_del(&lport->port_list);
-	if (nvme_fc_waiting_to_unload && list_empty(&nvme_fc_lport_list))
-		complete(&nvme_fc_unload_proceed);
 	spin_unlock_irqrestore(&nvme_fc_lock, flags);
 
 	ida_free(&nvme_fc_local_port_cnt, lport->localport.port_num);
@@ -3894,10 +3887,6 @@ static int __init nvme_fc_init_module(void)
 {
 	int ret;
 
-	nvme_fc_wq = alloc_workqueue("nvme_fc_wq", WQ_MEM_RECLAIM, 0);
-	if (!nvme_fc_wq)
-		return -ENOMEM;
-
 	/*
 	 * NOTE:
 	 * It is expected that in the future the kernel will combine
@@ -3915,7 +3904,7 @@ static int __init nvme_fc_init_module(void)
 	ret = class_register(&fc_class);
 	if (ret) {
 		pr_err("couldn't register class fc\n");
-		goto out_destroy_wq;
+		return ret;
 	}
 
 	/*
@@ -3939,8 +3928,6 @@ static int __init nvme_fc_init_module(void)
 	device_destroy(&fc_class, MKDEV(0, 0));
 out_destroy_class:
 	class_unregister(&fc_class);
-out_destroy_wq:
-	destroy_workqueue(nvme_fc_wq);
 
 	return ret;
 }
@@ -3960,45 +3947,23 @@ nvme_fc_delete_controllers(struct nvme_fc_rport *rport)
 	spin_unlock(&rport->lock);
 }
 
-static void
-nvme_fc_cleanup_for_unload(void)
+static void __exit nvme_fc_exit_module(void)
 {
 	struct nvme_fc_lport *lport;
 	struct nvme_fc_rport *rport;
-
-	list_for_each_entry(lport, &nvme_fc_lport_list, port_list) {
-		list_for_each_entry(rport, &lport->endp_list, endp_list) {
-			nvme_fc_delete_controllers(rport);
-		}
-	}
-}
-
-static void __exit nvme_fc_exit_module(void)
-{
 	unsigned long flags;
-	bool need_cleanup = false;
 
 	spin_lock_irqsave(&nvme_fc_lock, flags);
-	nvme_fc_waiting_to_unload = true;
-	if (!list_empty(&nvme_fc_lport_list)) {
-		need_cleanup = true;
-		nvme_fc_cleanup_for_unload();
-	}
+	list_for_each_entry(lport, &nvme_fc_lport_list, port_list)
+		list_for_each_entry(rport, &lport->endp_list, endp_list)
+			nvme_fc_delete_controllers(rport);
 	spin_unlock_irqrestore(&nvme_fc_lock, flags);
-	if (need_cleanup) {
-		pr_info("%s: waiting for ctlr deletes\n", __func__);
-		wait_for_completion(&nvme_fc_unload_proceed);
-		pr_info("%s: ctrl deletes complete\n", __func__);
-	}
+	flush_workqueue(nvme_delete_wq);
 
 	nvmf_unregister_transport(&nvme_fc_transport);
 
-	ida_destroy(&nvme_fc_local_port_cnt);
-	ida_destroy(&nvme_fc_ctrl_cnt);
-
 	device_destroy(&fc_class, MKDEV(0, 0));
 	class_unregister(&fc_class);
-	destroy_workqueue(nvme_fc_wq);
 }
 
 module_init(nvme_fc_init_module);
-- 
2.43.0

[PATCH v4 02/12] nvmet-fcloop: swap the list_add_tail arguments

[Daniel Wagner]

The first argument of list_add_tail function is the new element which
should be added to the list which is the second argument. Swap the
arguments to allow processing more than one element at a time.

Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fcloop.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/nvme/target/fcloop.c b/drivers/nvme/target/fcloop.c
index c1faeb1e9e55..1471af250ea6 100644
--- a/drivers/nvme/target/fcloop.c
+++ b/drivers/nvme/target/fcloop.c
@@ -358,7 +358,7 @@ fcloop_h2t_ls_req(struct nvme_fc_local_port *localport,
 	if (!rport->targetport) {
 		tls_req->status = -ECONNREFUSED;
 		spin_lock(&rport->lock);
-		list_add_tail(&rport->ls_list, &tls_req->ls_list);
+		list_add_tail(&tls_req->ls_list, &rport->ls_list);
 		spin_unlock(&rport->lock);
 		queue_work(nvmet_wq, &rport->ls_work);
 		return ret;
@@ -391,7 +391,7 @@ fcloop_h2t_xmt_ls_rsp(struct nvmet_fc_target_port *targetport,
 	if (remoteport) {
 		rport = remoteport->private;
 		spin_lock(&rport->lock);
-		list_add_tail(&rport->ls_list, &tls_req->ls_list);
+		list_add_tail(&tls_req->ls_list, &rport->ls_list);
 		spin_unlock(&rport->lock);
 		queue_work(nvmet_wq, &rport->ls_work);
 	}
@@ -446,7 +446,7 @@ fcloop_t2h_ls_req(struct nvmet_fc_target_port *targetport, void *hosthandle,
 	if (!tport->remoteport) {
 		tls_req->status = -ECONNREFUSED;
 		spin_lock(&tport->lock);
-		list_add_tail(&tport->ls_list, &tls_req->ls_list);
+		list_add_tail(&tls_req->ls_list, &tport->ls_list);
 		spin_unlock(&tport->lock);
 		queue_work(nvmet_wq, &tport->ls_work);
 		return ret;
-- 
2.43.0

[PATCH v4 03/12] nvmet-fc: release reference on target port

[Daniel Wagner]

In case we return early out of __nvmet_fc_finish_ls_req() we still have
to release the reference on the target port.

Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 05e6a755b330..3d53d9dc1099 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -360,7 +360,7 @@ __nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop)
 
 	if (!lsop->req_queued) {
 		spin_unlock_irqrestore(&tgtport->lock, flags);
-		return;
+		goto out_puttgtport;
 	}
 
 	list_del(&lsop->lsreq_list);
@@ -373,6 +373,7 @@ __nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop)
 				  (lsreq->rqstlen + lsreq->rsplen),
 				  DMA_BIDIRECTIONAL);
 
+out_puttgtport:
 	nvmet_fc_tgtport_put(tgtport);
 }
 
-- 
2.43.0

[PATCH v4 04/12] nvmet-fc: defer cleanup using RCU properly

[Daniel Wagner]

When the target executes a disconnect and the host triggers a reconnect
immediately, the reconnect command still finds an existing association.

The reconnect crashes later on because nvmet_fc_delete_target_assoc
blindly removes resources while the reconnect code wants to use it.

To address this, nvmet_fc_find_target_assoc should not be able to
lookup an association which is being removed. The association list
is already under RCU lifetime management, so let's properly use it
and remove the association from the list and wait for a grace period
before cleaning up all. This means we also can drop the RCU management
on the queues, because this is now handled via the association itself.

A second step split the execution context so that the initial disconnect
command can complete without running the reconnect code in the same
context. As usual, this is done by deferring the ->done to a workqueue.

Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 83 ++++++++++++++++++----------------------
 1 file changed, 37 insertions(+), 46 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 3d53d9dc1099..11ecfef41bd1 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -166,7 +166,7 @@ struct nvmet_fc_tgt_assoc {
 	struct nvmet_fc_hostport	*hostport;
 	struct nvmet_fc_ls_iod		*rcv_disconn;
 	struct list_head		a_list;
-	struct nvmet_fc_tgt_queue __rcu	*queues[NVMET_NR_QUEUES + 1];
+	struct nvmet_fc_tgt_queue 	*queues[NVMET_NR_QUEUES + 1];
 	struct kref			ref;
 	struct work_struct		del_work;
 	struct rcu_head			rcu;
@@ -803,14 +803,11 @@ nvmet_fc_alloc_target_queue(struct nvmet_fc_tgt_assoc *assoc,
 	if (!queue)
 		return NULL;
 
-	if (!nvmet_fc_tgt_a_get(assoc))
-		goto out_free_queue;
-
 	queue->work_q = alloc_workqueue("ntfc%d.%d.%d", 0, 0,
 				assoc->tgtport->fc_target_port.port_num,
 				assoc->a_id, qid);
 	if (!queue->work_q)
-		goto out_a_put;
+		goto out_free_queue;
 
 	queue->qid = qid;
 	queue->sqsize = sqsize;
@@ -832,15 +829,13 @@ nvmet_fc_alloc_target_queue(struct nvmet_fc_tgt_assoc *assoc,
 		goto out_fail_iodlist;
 
 	WARN_ON(assoc->queues[qid]);
-	rcu_assign_pointer(assoc->queues[qid], queue);
+	assoc->queues[qid] = queue;
 
 	return queue;
 
 out_fail_iodlist:
 	nvmet_fc_destroy_fcp_iodlist(assoc->tgtport, queue);
 	destroy_workqueue(queue->work_q);
-out_a_put:
-	nvmet_fc_tgt_a_put(assoc);
 out_free_queue:
 	kfree(queue);
 	return NULL;
@@ -853,12 +848,8 @@ nvmet_fc_tgt_queue_free(struct kref *ref)
 	struct nvmet_fc_tgt_queue *queue =
 		container_of(ref, struct nvmet_fc_tgt_queue, ref);
 
-	rcu_assign_pointer(queue->assoc->queues[queue->qid], NULL);
-
 	nvmet_fc_destroy_fcp_iodlist(queue->assoc->tgtport, queue);
 
-	nvmet_fc_tgt_a_put(queue->assoc);
-
 	destroy_workqueue(queue->work_q);
 
 	kfree_rcu(queue, rcu);
@@ -970,7 +961,7 @@ nvmet_fc_find_target_queue(struct nvmet_fc_tgtport *tgtport,
 	rcu_read_lock();
 	list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
 		if (association_id == assoc->association_id) {
-			queue = rcu_dereference(assoc->queues[qid]);
+			queue = assoc->queues[qid];
 			if (queue &&
 			    (!atomic_read(&queue->connected) ||
 			     !nvmet_fc_tgt_q_get(queue)))
@@ -1173,13 +1164,18 @@ nvmet_fc_target_assoc_free(struct kref *ref)
 	struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
 	struct nvmet_fc_ls_iod	*oldls;
 	unsigned long flags;
+	int i;
+
+	for (i = NVMET_NR_QUEUES; i >= 0; i--) {
+		if (assoc->queues[i])
+			nvmet_fc_delete_target_queue(assoc->queues[i]);
+	}
 
 	/* Send Disconnect now that all i/o has completed */
 	nvmet_fc_xmt_disconnect_assoc(assoc);
 
 	nvmet_fc_free_hostport(assoc->hostport);
 	spin_lock_irqsave(&tgtport->lock, flags);
-	list_del_rcu(&assoc->a_list);
 	oldls = assoc->rcv_disconn;
 	spin_unlock_irqrestore(&tgtport->lock, flags);
 	/* if pending Rcv Disconnect Association LS, send rsp now */
@@ -1209,7 +1205,7 @@ static void
 nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc)
 {
 	struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
-	struct nvmet_fc_tgt_queue *queue;
+	unsigned long flags;
 	int i, terminating;
 
 	terminating = atomic_xchg(&assoc->terminating, 1);
@@ -1218,29 +1214,21 @@ nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc)
 	if (terminating)
 		return;
 
+	spin_lock_irqsave(&tgtport->lock, flags);
+	list_del_rcu(&assoc->a_list);
+	spin_unlock_irqrestore(&tgtport->lock, flags);
 
-	for (i = NVMET_NR_QUEUES; i >= 0; i--) {
-		rcu_read_lock();
-		queue = rcu_dereference(assoc->queues[i]);
-		if (!queue) {
-			rcu_read_unlock();
-			continue;
-		}
+	synchronize_rcu();
 
-		if (!nvmet_fc_tgt_q_get(queue)) {
-			rcu_read_unlock();
-			continue;
-		}
-		rcu_read_unlock();
-		nvmet_fc_delete_target_queue(queue);
-		nvmet_fc_tgt_q_put(queue);
+	/* ensure all in-flight I/Os have been processed */
+	for (i = NVMET_NR_QUEUES; i >= 0; i--) {
+		if (assoc->queues[i])
+			flush_workqueue(assoc->queues[i]->work_q);
 	}
 
 	dev_info(tgtport->dev,
 		"{%d:%d} Association deleted\n",
 		tgtport->fc_target_port.port_num, assoc->a_id);
-
-	nvmet_fc_tgt_a_put(assoc);
 }
 
 static struct nvmet_fc_tgt_assoc *
@@ -1493,9 +1481,8 @@ __nvmet_fc_free_assocs(struct nvmet_fc_tgtport *tgtport)
 	list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
 		if (!nvmet_fc_tgt_a_get(assoc))
 			continue;
-		if (!queue_work(nvmet_wq, &assoc->del_work))
-			/* already deleting - release local reference */
-			nvmet_fc_tgt_a_put(assoc);
+		queue_work(nvmet_wq, &assoc->del_work);
+		nvmet_fc_tgt_a_put(assoc);
 	}
 	rcu_read_unlock();
 }
@@ -1548,9 +1535,8 @@ nvmet_fc_invalidate_host(struct nvmet_fc_target_port *target_port,
 			continue;
 		assoc->hostport->invalid = 1;
 		noassoc = false;
-		if (!queue_work(nvmet_wq, &assoc->del_work))
-			/* already deleting - release local reference */
-			nvmet_fc_tgt_a_put(assoc);
+		queue_work(nvmet_wq, &assoc->del_work);
+		nvmet_fc_tgt_a_put(assoc);
 	}
 	spin_unlock_irqrestore(&tgtport->lock, flags);
 
@@ -1582,7 +1568,7 @@ nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl)
 
 		rcu_read_lock();
 		list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
-			queue = rcu_dereference(assoc->queues[0]);
+			queue = assoc->queues[0];
 			if (queue && queue->nvme_sq.ctrl == ctrl) {
 				if (nvmet_fc_tgt_a_get(assoc))
 					found_ctrl = true;
@@ -1594,9 +1580,8 @@ nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl)
 		nvmet_fc_tgtport_put(tgtport);
 
 		if (found_ctrl) {
-			if (!queue_work(nvmet_wq, &assoc->del_work))
-				/* already deleting - release local reference */
-				nvmet_fc_tgt_a_put(assoc);
+			queue_work(nvmet_wq, &assoc->del_work);
+			nvmet_fc_tgt_a_put(assoc);
 			return;
 		}
 
@@ -1626,6 +1611,8 @@ nvmet_fc_unregister_targetport(struct nvmet_fc_target_port *target_port)
 	/* terminate any outstanding associations */
 	__nvmet_fc_free_assocs(tgtport);
 
+	flush_workqueue(nvmet_wq);
+
 	/*
 	 * should terminate LS's as well. However, LS's will be generated
 	 * at the tail end of association termination, so they likely don't
@@ -1871,9 +1858,6 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport,
 				sizeof(struct fcnvme_ls_disconnect_assoc_acc)),
 			FCNVME_LS_DISCONNECT_ASSOC);
 
-	/* release get taken in nvmet_fc_find_target_assoc */
-	nvmet_fc_tgt_a_put(assoc);
-
 	/*
 	 * The rules for LS response says the response cannot
 	 * go back until ABTS's have been sent for all outstanding
@@ -1888,8 +1872,6 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport,
 	assoc->rcv_disconn = iod;
 	spin_unlock_irqrestore(&tgtport->lock, flags);
 
-	nvmet_fc_delete_target_assoc(assoc);
-
 	if (oldls) {
 		dev_info(tgtport->dev,
 			"{%d:%d} Multiple Disconnect Association LS's "
@@ -1905,6 +1887,9 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport,
 		nvmet_fc_xmt_ls_rsp(tgtport, oldls);
 	}
 
+	queue_work(nvmet_wq, &assoc->del_work);
+	nvmet_fc_tgt_a_put(assoc);
+
 	return false;
 }
 
@@ -2903,6 +2888,9 @@ nvmet_fc_remove_port(struct nvmet_port *port)
 
 	nvmet_fc_portentry_unbind(pe);
 
+	/* terminate any outstanding associations */
+	__nvmet_fc_free_assocs(pe->tgtport);
+
 	kfree(pe);
 }
 
@@ -2934,6 +2922,9 @@ static int __init nvmet_fc_init_module(void)
 
 static void __exit nvmet_fc_exit_module(void)
 {
+	/* ensure any shutdown operation, e.g. delete ctrls have finished */
+	flush_workqueue(nvmet_wq);
+
 	/* sanity check - all lports should be removed */
 	if (!list_empty(&nvmet_fc_target_list))
 		pr_warn("%s: targetport list not empty\n", __func__);
-- 
2.43.0

[PATCH v4 05/12] nvmet-fc: free queue and assoc directly

[Daniel Wagner]

Neither struct nvmet_fc_tgt_queue nor struct nvmet_fc_tgt_assoc are data
structure which are used in a RCU context. So there is no reason to
delay the free operation.

Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 11ecfef41bd1..b44b99525c44 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -145,7 +145,6 @@ struct nvmet_fc_tgt_queue {
 	struct list_head		avail_defer_list;
 	struct workqueue_struct		*work_q;
 	struct kref			ref;
-	struct rcu_head			rcu;
 	/* array of fcp_iods */
 	struct nvmet_fc_fcp_iod		fod[] __counted_by(sqsize);
 } __aligned(sizeof(unsigned long long));
@@ -169,7 +168,6 @@ struct nvmet_fc_tgt_assoc {
 	struct nvmet_fc_tgt_queue 	*queues[NVMET_NR_QUEUES + 1];
 	struct kref			ref;
 	struct work_struct		del_work;
-	struct rcu_head			rcu;
 };
 
 
@@ -852,7 +850,7 @@ nvmet_fc_tgt_queue_free(struct kref *ref)
 
 	destroy_workqueue(queue->work_q);
 
-	kfree_rcu(queue, rcu);
+	kfree(queue);
 }
 
 static void
@@ -1185,8 +1183,8 @@ nvmet_fc_target_assoc_free(struct kref *ref)
 	dev_info(tgtport->dev,
 		"{%d:%d} Association freed\n",
 		tgtport->fc_target_port.port_num, assoc->a_id);
-	kfree_rcu(assoc, rcu);
 	nvmet_fc_tgtport_put(tgtport);
+	kfree(assoc);
 }
 
 static void
-- 
2.43.0

[PATCH v4 06/12] nvmet-fc: hold reference on hostport match

[Daniel Wagner]

The hostport data structure is shared between the association, this why
we keep track of the users via a refcount. So we should not decrement
the refcount on a match and free the hostport several times.

Reported by KASAN.

Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index b44b99525c44..205a12b1e841 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -1068,8 +1068,6 @@ nvmet_fc_alloc_hostport(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 		/* new allocation not needed */
 		kfree(newhost);
 		newhost = match;
-		/* no new allocation - release reference */
-		nvmet_fc_tgtport_put(tgtport);
 	} else {
 		newhost->tgtport = tgtport;
 		newhost->hosthandle = hosthandle;
-- 
2.43.0

[PATCH v4 07/12] nvmet-fc: remove null hostport pointer check

[Daniel Wagner]

An association has always a valid hostport pointer. Remove useless
null pointer check.

Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 205a12b1e841..849d9b17c60a 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -488,8 +488,7 @@ nvmet_fc_xmt_disconnect_assoc(struct nvmet_fc_tgt_assoc *assoc)
 	 * message is normal. Otherwise, send unless the hostport has
 	 * already been invalidated by the lldd.
 	 */
-	if (!tgtport->ops->ls_req || !assoc->hostport ||
-	    assoc->hostport->invalid)
+	if (!tgtport->ops->ls_req || assoc->hostport->invalid)
 		return;
 
 	lsop = kzalloc((sizeof(*lsop) +
@@ -1524,8 +1523,7 @@ nvmet_fc_invalidate_host(struct nvmet_fc_target_port *target_port,
 	spin_lock_irqsave(&tgtport->lock, flags);
 	list_for_each_entry_safe(assoc, next,
 				&tgtport->assoc_list, a_list) {
-		if (!assoc->hostport ||
-		    assoc->hostport->hosthandle != hosthandle)
+		if (assoc->hostport->hosthandle != hosthandle)
 			continue;
 		if (!nvmet_fc_tgt_a_get(assoc))
 			continue;
-- 
2.43.0

[PATCH v4 08/12] nvmet-fc: do not tack refs on tgtports from assoc

[Daniel Wagner]

The association life time is tied to the life time of the target port.
That means we should not take extra a refcount when creating a
association.

Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 849d9b17c60a..fe3246024836 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -1107,12 +1107,9 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 	if (idx < 0)
 		goto out_free_assoc;
 
-	if (!nvmet_fc_tgtport_get(tgtport))
-		goto out_ida;
-
 	assoc->hostport = nvmet_fc_alloc_hostport(tgtport, hosthandle);
 	if (IS_ERR(assoc->hostport))
-		goto out_put;
+		goto out_ida;
 
 	assoc->tgtport = tgtport;
 	assoc->a_id = idx;
@@ -1142,8 +1139,6 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 
 	return assoc;
 
-out_put:
-	nvmet_fc_tgtport_put(tgtport);
 out_ida:
 	ida_free(&tgtport->assoc_cnt, idx);
 out_free_assoc:
@@ -1180,7 +1175,6 @@ nvmet_fc_target_assoc_free(struct kref *ref)
 	dev_info(tgtport->dev,
 		"{%d:%d} Association freed\n",
 		tgtport->fc_target_port.port_num, assoc->a_id);
-	nvmet_fc_tgtport_put(tgtport);
 	kfree(assoc);
 }
 
-- 
2.43.0

[PATCH v4 09/12] nvmet-fc: abort command when there is no binding

[Daniel Wagner]

When the target port has not active port binding, there is no point in
trying to process the command as it has to fail anyway. Instead adding
checks to all commands abort the command early.

Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index fe3246024836..c80b8a066fd1 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -1099,6 +1099,9 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 	int idx;
 	bool needrandom = true;
 
+	if (!tgtport->pe)
+		return NULL;
+
 	assoc = kzalloc(sizeof(*assoc), GFP_KERNEL);
 	if (!assoc)
 		return NULL;
@@ -2514,8 +2517,9 @@ nvmet_fc_handle_fcp_rqst(struct nvmet_fc_tgtport *tgtport,
 
 	fod->req.cmd = &fod->cmdiubuf.sqe;
 	fod->req.cqe = &fod->rspiubuf.cqe;
-	if (tgtport->pe)
-		fod->req.port = tgtport->pe->port;
+	if (!tgtport->pe)
+		goto transport_error;
+	fod->req.port = tgtport->pe->port;
 
 	/* clear any response payload */
 	memset(&fod->rspiubuf, 0, sizeof(fod->rspiubuf));
-- 
2.43.0

[PATCH v4 10/12] nvmet-fc: avoid deadlock on delete association path

[Daniel Wagner]

When deleting an association the shutdown path is deadlocking because we
try to flush the nvmet_wq nested. Avoid this by deadlock by deferring
the put work into its own work item.

Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index c80b8a066fd1..3e0d391e631b 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -111,6 +111,8 @@ struct nvmet_fc_tgtport {
 	struct nvmet_fc_port_entry	*pe;
 	struct kref			ref;
 	u32				max_sg_cnt;
+
+	struct work_struct		put_work;
 };
 
 struct nvmet_fc_port_entry {
@@ -247,6 +249,13 @@ static int nvmet_fc_tgt_a_get(struct nvmet_fc_tgt_assoc *assoc);
 static void nvmet_fc_tgt_q_put(struct nvmet_fc_tgt_queue *queue);
 static int nvmet_fc_tgt_q_get(struct nvmet_fc_tgt_queue *queue);
 static void nvmet_fc_tgtport_put(struct nvmet_fc_tgtport *tgtport);
+static void nvmet_fc_put_tgtport_work(struct work_struct *work)
+{
+	struct nvmet_fc_tgtport *tgtport =
+		container_of(work, struct nvmet_fc_tgtport, put_work);
+
+	nvmet_fc_tgtport_put(tgtport);
+}
 static int nvmet_fc_tgtport_get(struct nvmet_fc_tgtport *tgtport);
 static void nvmet_fc_handle_fcp_rqst(struct nvmet_fc_tgtport *tgtport,
 					struct nvmet_fc_fcp_iod *fod);
@@ -358,7 +367,7 @@ __nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop)
 
 	if (!lsop->req_queued) {
 		spin_unlock_irqrestore(&tgtport->lock, flags);
-		goto out_puttgtport;
+		goto out_putwork;
 	}
 
 	list_del(&lsop->lsreq_list);
@@ -371,8 +380,8 @@ __nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop)
 				  (lsreq->rqstlen + lsreq->rsplen),
 				  DMA_BIDIRECTIONAL);
 
-out_puttgtport:
-	nvmet_fc_tgtport_put(tgtport);
+out_putwork:
+	queue_work(nvmet_wq, &tgtport->put_work);
 }
 
 static int
@@ -1396,6 +1405,7 @@ nvmet_fc_register_targetport(struct nvmet_fc_port_info *pinfo,
 	kref_init(&newrec->ref);
 	ida_init(&newrec->assoc_cnt);
 	newrec->max_sg_cnt = template->max_sgl_segments;
+	INIT_WORK(&newrec->put_work, nvmet_fc_put_tgtport_work);
 
 	ret = nvmet_fc_alloc_ls_iodlist(newrec);
 	if (ret) {
-- 
2.43.0

[PATCH v4 11/12] nvmet-fc: take ref count on tgtport before delete assoc

[Daniel Wagner]

We have to ensure that the tgtport is not going away
before be have remove all the associations.

Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 31 +++++++++++++++++++++++--------
 1 file changed, 23 insertions(+), 8 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 3e0d391e631b..671d096745a5 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -1090,13 +1090,28 @@ nvmet_fc_alloc_hostport(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 }
 
 static void
-nvmet_fc_delete_assoc(struct work_struct *work)
+nvmet_fc_delete_assoc(struct nvmet_fc_tgt_assoc *assoc)
+{
+	nvmet_fc_delete_target_assoc(assoc);
+	nvmet_fc_tgt_a_put(assoc);
+}
+
+static void
+nvmet_fc_delete_assoc_work(struct work_struct *work)
 {
 	struct nvmet_fc_tgt_assoc *assoc =
 		container_of(work, struct nvmet_fc_tgt_assoc, del_work);
+	struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
 
-	nvmet_fc_delete_target_assoc(assoc);
-	nvmet_fc_tgt_a_put(assoc);
+	nvmet_fc_delete_assoc(assoc);
+	nvmet_fc_tgtport_put(tgtport);
+}
+
+static void
+nvmet_fc_schedule_delete_assoc(struct nvmet_fc_tgt_assoc *assoc)
+{
+	nvmet_fc_tgtport_get(assoc->tgtport);
+	queue_work(nvmet_wq, &assoc->del_work);
 }
 
 static struct nvmet_fc_tgt_assoc *
@@ -1127,7 +1142,7 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 	assoc->a_id = idx;
 	INIT_LIST_HEAD(&assoc->a_list);
 	kref_init(&assoc->ref);
-	INIT_WORK(&assoc->del_work, nvmet_fc_delete_assoc);
+	INIT_WORK(&assoc->del_work, nvmet_fc_delete_assoc_work);
 	atomic_set(&assoc->terminating, 0);
 
 	while (needrandom) {
@@ -1483,7 +1498,7 @@ __nvmet_fc_free_assocs(struct nvmet_fc_tgtport *tgtport)
 	list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
 		if (!nvmet_fc_tgt_a_get(assoc))
 			continue;
-		queue_work(nvmet_wq, &assoc->del_work);
+		nvmet_fc_schedule_delete_assoc(assoc);
 		nvmet_fc_tgt_a_put(assoc);
 	}
 	rcu_read_unlock();
@@ -1536,7 +1551,7 @@ nvmet_fc_invalidate_host(struct nvmet_fc_target_port *target_port,
 			continue;
 		assoc->hostport->invalid = 1;
 		noassoc = false;
-		queue_work(nvmet_wq, &assoc->del_work);
+		nvmet_fc_schedule_delete_assoc(assoc);
 		nvmet_fc_tgt_a_put(assoc);
 	}
 	spin_unlock_irqrestore(&tgtport->lock, flags);
@@ -1581,7 +1596,7 @@ nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl)
 		nvmet_fc_tgtport_put(tgtport);
 
 		if (found_ctrl) {
-			queue_work(nvmet_wq, &assoc->del_work);
+			nvmet_fc_schedule_delete_assoc(assoc);
 			nvmet_fc_tgt_a_put(assoc);
 			return;
 		}
@@ -1888,7 +1903,7 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport,
 		nvmet_fc_xmt_ls_rsp(tgtport, oldls);
 	}
 
-	queue_work(nvmet_wq, &assoc->del_work);
+	nvmet_fc_schedule_delete_assoc(assoc);
 	nvmet_fc_tgt_a_put(assoc);
 
 	return false;
-- 
2.43.0

[PATCH v4 12/12] nvmet-fc: use RCU list iterator for assoc_list

[Daniel Wagner]

The assoc_list is a RCU protected list, thus use the RCU flavor of list
functions.

Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/target/fc.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 671d096745a5..b8d3e30f2ddb 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -1151,7 +1151,9 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 
 		spin_lock_irqsave(&tgtport->lock, flags);
 		needrandom = false;
-		list_for_each_entry(tmpassoc, &tgtport->assoc_list, a_list) {
+		rcu_read_lock();
+		list_for_each_entry_rcu(tmpassoc, &tgtport->assoc_list, a_list) {
+
 			if (ran == tmpassoc->association_id) {
 				needrandom = true;
 				break;
@@ -1161,6 +1163,7 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
 			assoc->association_id = ran;
 			list_add_tail_rcu(&assoc->a_list, &tgtport->assoc_list);
 		}
+		rcu_read_unlock();
 		spin_unlock_irqrestore(&tgtport->lock, flags);
 	}
 
-- 
2.43.0

Re: [PATCH v4 02/12] nvmet-fcloop: swap the list_add_tail arguments

[Christoph Hellwig]

On Tue, Jan 30, 2024 at 10:49:28AM +0100, Daniel Wagner wrote:
> The first argument of list_add_tail function is the new element which
> should be added to the list which is the second argument. Swap the
> arguments to allow processing more than one element at a time.

Ooops.

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 03/12] nvmet-fc: release reference on target port

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 04/12] nvmet-fc: defer cleanup using RCU properly

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 05/12] nvmet-fc: free queue and assoc directly

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 06/12] nvmet-fc: hold reference on hostport match

[Christoph Hellwig]

On Tue, Jan 30, 2024 at 10:49:32AM +0100, Daniel Wagner wrote:
> The hostport data structure is shared between the association, this why
> we keep track of the users via a refcount. So we should not decrement
> the refcount on a match and free the hostport several times.
> 
> Reported by KASAN.
> 
> Reviewed-by: Hannes Reinecke <hare@suse.de>
> Signed-off-by: Daniel Wagner <dwagner@suse.de>

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 07/12] nvmet-fc: remove null hostport pointer check

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 08/12] nvmet-fc: do not tack refs on tgtports from assoc

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 09/12] nvmet-fc: abort command when there is no binding

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 10/12] nvmet-fc: avoid deadlock on delete association path

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 11/12] nvmet-fc: take ref count on tgtport before delete assoc

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 12/12] nvmet-fc: use RCU list iterator for assoc_list

[Christoph Hellwig]

> +		list_for_each_entry_rcu(tmpassoc, &tgtport->assoc_list, a_list) {

please avoid the overly long line.

Also in general the (rcu)lock, list walk to find something, return
patterns tend to benefit from being factored into a little helper.
Not sure if that really fits here, though.